Personalized Neoantigen Peptide Immunogenicity Correlates with Hydrophobicity and HLA Binding in Cancer Patients

Despite the promise of personalized neoantigen vaccines in inducing antitumor T cell responses, a significant challenge remains: only 10-20% of selected peptides elicit immune responses in patients. This low success rate underscores critical limitations in current neoantigen prediction strategies, hindering the development of effective cancer immunotherapies. A deeper understanding of the intrinsic peptide characteristics and antigen-processing machinery interactions that drive immunogenicity is crucial to improve the design and efficacy of these precision medicine approaches, particularly for solid tumors like non-small cell lung cancer (NSCLC) where current checkpoint inhibitors have limited impact.

Researchers analyzed a clinically annotated dataset from 352 cancer patients who received personalized neoantigen peptide-pulsed dendritic cell vaccines. The study focused on 2,317 short peptides derived from single nucleotide variants. Post-vaccination T cell responses were rigorously evaluated using IFN-γ ELISPOT assay. Immunogenic neoantigen peptides were specifically defined as those inducing a ≥2.0-fold increase in IFN-γ ELISPOT responses after vaccination. The team systematically examined peptide intrinsic characteristics, physicochemical properties, and various predicted scores related to the antigen-processing machinery to identify determinants of immunogenicity.

Immunogenicity of neoantigen peptides was not associated with specific mutation positions or sequence patterns. However, a significant correlation was found with higher hydrophobicity (P = 5.2 × 10-4). Several predictive scores were significantly enriched among immunogenic neoantigen peptides, indicating their importance in T cell activation. Peptides demonstrating higher binding affinity to HLA molecules showed strong correlations (P = 0.0014 for NetMHC3, P = 0.028 for MHCflurry-affinity). Similarly, higher binding stability (P = 0.043 for NetMHCstab) and better peptide presentation scores (P = 0.012 for mixmhcPred3, P = 0.0085 for MHCflurry-presentation) were also significantly linked to immunogenicity. Composite models, which integrated peptide physicochemical features—especially hydrophobicity—with these prediction scores, demonstrably improved the area under the receiver operating characteristic curve (AUC) and balanced accuracy compared to using individual tools alone. This highlights the multifactorial nature of neoantigen immunogenicity.

Immunogenic neoantigen peptides were significantly enriched with higher hydrophobicity, better HLA binding affinity, and improved peptide presentation scores.